Upload folder using huggingface_hub

README.md

@@ -24,76 +24,122 @@ optimised for training a 500M parameter model focused on structured data output
 
 - **Total code files:** 5,203,508
 - **Total tokens:** 3.9B (target: 3.5B)
+- **Classifier-scored files:** 1,553,596 (1.7B tokens)
+- **Non-classified files:** 3,649,912 (2.2B tokens) — filtered by heuristics, not the classifier
 - **Source:** bigcode/starcoderdata
 - **Classifier:** [mdonigian/code-curator-v1](https://huggingface.co/mdonigian/code-curator-v1) (UniXcoder-base, multi-task)
 - **Curation:** Per-language-slice filtering + compression ratio pre-filter + MinHash deduplication
 
 ## Filtering Strategy
 
-Each language slice has its own filtering strategy and token budget:
+Different language groups need different curation approaches. Not every slice
+goes through the GPU classifier — schema languages and GitHub issues are filtered
+with cheaper heuristics because the classifier was trained on general-purpose code
+and isn't the right tool for inherently structured formats.
 
-1. **Schema languages** (JSON/YAML/SQL/protobuf/GraphQL): ~800M tokens, light filter (quality floor only)
-2. **TypeScript** (relevance ≥ 2): ~600M tokens, relevance classifier
-3. **Python** (relevance ≥ 2): ~600M tokens, relevance classifier
-4. **Rust/Go/Java** (relevance ≥ 2): ~600M tokens, relevance classifier
-5. **Jupyter notebooks**: ~400M tokens, passthrough (already structured)
-6. **GitHub Issues** (technical): ~500M tokens, keyword filter
+**All slices** share these pre-filters:
+- zlib compression ratio < 0.10 (catches extreme repetition)
+- MinHash LSH deduplication (128 perms, 5-line shingles, 0.7 Jaccard threshold)
 
-Pre-filtering applied to all slices:
-- zlib compression ratio < 0.10 catches extreme repetition
-- Quality floor: drop files with predicted quality ≤ 1 (broken/gibberish)
-- MinHash LSH deduplication with line-level 5-gram shingles
+### Classifier-Scored Slices (relevance_filter)
+
+These languages were scored by the multi-task classifier. Files were ranked by
+structured data relevance and filtered to keep only those with relevance ≥ 2.0
+and quality ≥ 1.5, sampled down to the per-slice token budget:
+
+- **TypeScript**: ~600M tokens — strong type system, filter by SD relevance ≥ 2
+- **Python**: ~600M tokens — filter by SD relevance ≥ 2
+- **Rust/Go/Java**: ~600M tokens — strongly typed, filter by SD relevance ≥ 2
+- **Jupyter notebooks**: ~400M tokens — filter by SD relevance ≥ 2
+
+### Non-Classified Slices
+
+These languages were **not** run through the classifier. Their `quality`,
+`structured_data`, and `content_type` columns contain default placeholder values
+(0.0 / "unclassified") and should be ignored:
+
+- **Schema languages** (JSON/YAML/SQL/protobuf/thrift/XSLT): ~800M tokens — inherently structured data formats; quality floor + random sample to budget
+- **GitHub Issues** (technical): ~500M tokens — keyword filter matching structured-data topics (JSON, schema, API, protobuf, gRPC, etc.)
+- **General code** (78 other languages): ~1B tokens — random sample for language diversity; quality floor only
 
 ## Language Slice Distribution
 
-| Slice | Strategy | Target | Actual | % |
-|-------|----------|--------|--------|---|
-| schema_languages | light_filter | 800M | 799M | 99.9% |
-| typescript | relevance_filter | 600M | 598M | 99.7% |
-| python | relevance_filter | 600M | 594M | 99.1% |
-| rust_go_java | relevance_filter | 600M | 485M | 80.8% |
-| jupyter | relevance_filter | 400M | 0M | 0.0% |
-| github_issues | keyword_filter | 500M | 426M | 85.2% |
-| general_code | light_filter | 1000M | 999M | 99.9% |
+| Slice | Strategy | Languages | Target | Actual | % of Target |
+|-------|----------|-----------|--------|--------|-------------|
+| schema_languages | light_filter | json, yaml, sql, protocol-buffer +2 more | 800M | 799M | 99.9% |
+| typescript | relevance_filter | typescript | 600M | 598M | 99.7% |
+| python | relevance_filter | python | 600M | 594M | 99.1% |
+| rust_go_java | relevance_filter | rust, go, java | 600M | 485M | 80.8% |
+| jupyter | relevance_filter | jupyter-scripts-dedup-filtered | 400M | 0M | 0.0% |
+| github_issues | keyword_filter | github-issues-filtered-structured | 500M | 426M | 85.2% |
+| general_code | light_filter | ada, agda, alloy, antlr +74 more | 1000M | 999M | 99.9% |
+
+
+## Classifier-Scored Slices — Detail
+
+The quality and structured data scores below apply **only** to the 1,553,596 files
+(1.7B tokens) that went through the classifier. Non-classified slices
+are excluded from these statistics.
 
+| Slice | Files | Tokens | Avg Quality | Avg SD Relevance |
+|-------|-------|--------|-------------|------------------|
+| typescript | 841,426 | 598M | 3.81 | 2.88 |
+| python | 567,721 | 594M | 3.71 | 2.73 |
+| rust_go_java | 144,438 | 485M | 3.97 | 3.07 |
+| jupyter | 11 | 0M | 3.11 | 2.23 |
 
-## Content Group Distribution
 
-| Group | % of Tokens | Tokens | Files |
-|-------|-------------|--------|-------|
-| Library/Package | 27.7% | 1,079,381,502 | 1,075,672 |
-| Application | 1.4% | 56,360,750 | 118,200 |
-| Script/CLI | 0.5% | 17,853,662 | 24,871 |
-| Test Code | 2.3% | 91,370,763 | 48,003 |
-| Config/Data/Generated/Other | 68.1% | 2,656,726,472 | 3,936,762 |
+### Content Group Distribution (classifier-scored files only)
 
+| Group | % of Classified Tokens | Tokens | Files |
+|-------|-----------------------|--------|-------|
+| Library/Package | 64.3% | 1,079,381,502 | 1,075,672 |
+| Application | 3.4% | 56,360,750 | 118,200 |
+| Script/CLI | 1.1% | 17,853,662 | 24,871 |
+| Test Code | 5.5% | 91,370,763 | 48,003 |
+| Config/Data/Generated/Other | 25.8% | 432,393,146 | 286,850 |
 
-## Structured Data Relevance Distribution
 
-The strongest classifier signal (Spearman 0.81). SD2+ files contain significant
-structured data patterns (API endpoints, JSON parsing, schema definitions, etc.).
+### Structured Data Relevance (classifier-scored files only)
+
+The strongest classifier signal (Spearman 0.81 on held-out data). SD2+ files
+contain significant structured data patterns (API endpoints, JSON parsing,
+schema definitions, etc.).
+
+Quality mean: 3.79, Median: 3.88.
 
 | Level | Range | Target % | Actual % | Files |
 |-------|-------|----------|----------|-------|
-| SD0 | [0.0, 0.5) | 10.0% | 70.1% | 3,649,912 |
+| SD0 | [0.0, 0.5) | 10.0% | 0.0% | 0 |
 | SD1 | [0.5, 1.5) | 20.0% | 0.0% | 0 |
-| SD2 | [1.5, 2.5) | 35.0% | 0.9% | 49,213 |
-| SD3 | [2.5, 3.5) | 35.0% | 28.9% | 1,504,383 |
+| SD2 | [1.5, 2.5) | 35.0% | 3.2% | 49,213 |
+| SD3 | [2.5, 3.5) | 35.0% | 96.8% | 1,504,383 |
 
 
-## Quality Distribution
-
-Quality mean: 1.13, Median: 0.00.
+### Quality Distribution (classifier-scored files only)
 
 | Level | Description | Files |
 |-------|-------------|-------|
-| 1 | Broken/gibberish | 3,649,912 |
+| 1 | Broken/gibberish | 0 |
 | 2 | Functional but poor | 42,668 |
 | 3 | Decent | 129,945 |
 | 4 | Good | 1,380,674 |
 | 5 | Excellent | 309 |
 
 
+## Non-Classified Slices — Detail
+
+These slices were filtered using heuristics. The classifier columns (`quality`,
+`structured_data`, `content_type`) are set to defaults and **do not reflect
+actual code quality** — the filtering was done by other means:
+
+| Slice | Strategy | Files | Tokens | How Filtered |
+|-------|----------|-------|--------|-------------|
+| schema_languages | light_filter | 2,203,233 | 799M | Quality floor (≥1.5) + token budget, randomly sampled |
+| github_issues | keyword_filter | 485,384 | 426M | Keyword match for structured-data topics + quality floor |
+| general_code | light_filter | 961,295 | 999M | Quality floor (≥1.5) + token budget, randomly sampled |
+
+
 ## Programming Languages
 
 | Language | % Tokens | Files |
@@ -136,19 +182,25 @@ Each row contains:
 | `lang` | string | Programming language |
 | `size` | int | File size in bytes |
 | `token_count` | int | Estimated token count (size // 4) |
-| `quality` | float | Code quality score (1-5) |
-| `structured_data` | float | Structured data relevance (0-3) |
-| `content_type` | string | Content type (9 classes) |
-| `language_slice` | string | Language slice (schema_languages, typescript, python, etc.) |
-| `relevance_score` | float | Composite relevance score |
+| `quality` | float | Code quality score 1-5 (**classifier-scored slices only**; 0.0 for non-classified) |
+| `structured_data` | float | Structured data relevance 0-3 (**classifier-scored slices only**; 0.0 for non-classified) |
+| `content_type` | string | Content type — 9 classes (**classifier-scored slices only**; "unclassified" for non-classified) |
+| `language_slice` | string | Language slice name (use this to filter by curation strategy) |
+| `relevance_score` | float | Composite relevance score (**classifier-scored slices only**; 0.0 for non-classified) |
+
+> **Tip:** To work with only classifier-scored data, filter on
+> `language_slice` in `{"typescript", "python", "rust_go_java", "jupyter"}`.
 
 ## Methodology
 
-1. **Classification:** Single multi-task UniXcoder-base model with three heads runs
-   on each code file. Stores raw float scores for maximum flexibility.
-2. **Pre-filtering:** zlib compression ratio filter removes repetitive boilerplate
+1. **Download:** All language folders from `bigcode/starcoderdata`.
+2. **Classification:** Multi-task UniXcoder-base model (3 heads: quality, SD relevance,
+   content type) runs on TypeScript, Python, Rust, Go, Java, and Jupyter files.
+   Schema languages, GitHub issues, and general code skip this step.
+3. **Pre-filtering:** zlib compression ratio filter removes repetitive boilerplate
    before GPU inference.
-3. **Filtering:** Quality floor + priority-based sampling targeting high structured
-   data relevance and balanced content type distribution.
-4. **Deduplication:** MinHash LSH (128 perms, 5-line shingles, 0.7 Jaccard threshold).
-   Highest-relevance code file kept from each cluster.
+4. **Filtering:** Per-slice strategy — relevance-based ranking for classified languages,
+   keyword matching for GitHub issues, random sampling for schema/general code. All
+   slices enforce a quality floor.
+5. **Deduplication:** MinHash LSH (128 perms, 5-line shingles, 0.7 Jaccard threshold).
+   Highest-relevance file kept from each cluster.

dataset_stats.json

@@ -2,41 +2,38 @@
   "total_documents": 5203508,
   "total_tokens": 3901693149,
   "num_output_files": 6,
+  "classified_documents": 1553596,
+  "classified_tokens": 1677359823,
   "group_distribution": {
     "library": {
       "display_name": "Library/Package",
       "docs": 1075672,
       "tokens": 1079381502,
-      "pct_tokens": 27.66,
-      "pct_of_total": 27.66
+      "pct_tokens": 64.35
     },
     "application": {
       "display_name": "Application",
       "docs": 118200,
       "tokens": 56360750,
-      "pct_tokens": 1.44,
-      "pct_of_total": 1.44
+      "pct_tokens": 3.36
     },
     "script": {
       "display_name": "Script/CLI",
       "docs": 24871,
       "tokens": 17853662,
-      "pct_tokens": 0.46,
-      "pct_of_total": 0.46
+      "pct_tokens": 1.06
     },
     "test": {
       "display_name": "Test Code",
       "docs": 48003,
       "tokens": 91370763,
-      "pct_tokens": 2.34,
-      "pct_of_total": 2.34
+      "pct_tokens": 5.45
     },
     "low_value": {
       "display_name": "Config/Data/Generated/Other",
-      "docs": 3936762,
-      "tokens": 2656726472,
-      "pct_tokens": 68.09,
-      "pct_of_total": 68.09
+      "docs": 286850,
+      "tokens": 432393146,
+      "pct_tokens": 25.78
     }
   },
   "content_type_distribution": {
@@ -51,23 +48,23 @@
     "other": 0
   },
   "quality_distribution": {
-    "1": 3649912,
+    "1": 0,
     "2": 42668,
     "3": 129945,
     "4": 1380674,
     "5": 309
   },
   "quality_stats": {
-    "mean": 1.131,
-    "std": 1.75,
-    "median": 0.0
+    "mean": 3.788,
+    "std": 0.432,
+    "median": 3.875
   },
   "sd_distribution": {
     "SD0": {
       "range": "[0.0, 0.5)",
-      "docs": 3649912,
-      "tokens": 2224333326,
-      "pct": 70.14,
+      "docs": 0,
+      "tokens": 0,
+      "pct": 0.0,
       "target_pct": 10.0
     },
     "SD1": {
@@ -81,21 +78,21 @@
       "range": "[1.5, 2.5)",
       "docs": 49213,
       "tokens": 55140562,
-      "pct": 0.95,
+      "pct": 3.17,
       "target_pct": 35.0
     },
     "SD3": {
       "range": "[2.5, 3.5)",
       "docs": 1504383,
       "tokens": 1622219261,
-      "pct": 28.91,
+      "pct": 96.83,
       "target_pct": 35.0
     }
   },
   "sd_stats": {
-    "mean": 0.849,
-    "std": 1.306,
-    "median": 0.0
+    "mean": 2.844,
+    "std": 0.19,
+    "median": 2.844
   },
   "top_languages": [
     {
@@ -223,7 +220,9 @@
       "actual_tokens": 598255831,
       "docs": 841426,
       "pct_of_target": 99.7,
-      "pct_of_total": 15.3
+      "pct_of_total": 15.3,
+      "quality_mean": 3.81,
+      "sd_mean": 2.88
     },
     "python": {
       "description": "Python \u2014 filter by structured data relevance \u2265 2",
@@ -235,7 +234,9 @@
       "actual_tokens": 594491555,
       "docs": 567721,
       "pct_of_target": 99.1,
-      "pct_of_total": 15.2
+      "pct_of_total": 15.2,
+      "quality_mean": 3.71,
+      "sd_mean": 2.73
     },
     "rust_go_java": {
       "description": "Rust/Go/Java \u2014 strongly typed, filter by structured data relevance \u2265 2",
@@ -249,7 +250,9 @@
       "actual_tokens": 484584034,
       "docs": 144438,
       "pct_of_target": 80.8,
-      "pct_of_total": 12.4
+      "pct_of_total": 12.4,
+      "quality_mean": 3.97,
+      "sd_mean": 3.07
     },
     "jupyter": {
       "description": "Jupyter notebooks \u2014 filter by structured data relevance \u2265 2",
@@ -261,7 +264,9 @@
       "actual_tokens": 28403,
       "docs": 11,
       "pct_of_target": 0.0,
-      "pct_of_total": 0.0
+      "pct_of_total": 0.0,
+      "quality_mean": 3.11,
+      "sd_mean": 2.23
     },
     "github_issues": {
       "description": "GitHub issues (technical) \u2014 keyword filter for structured data topics",